Ultra-robust Y-doped hafnium oxide ferroelectric memristors for intelligent edge computing
The rapid development of artificial intelligence (AI) and big data-driven edge intelligence applications has created an urgent demand for highly efficient computing hardware. Ferroelectric memristors have emerged as promising candidates for edge hardware due to their multi-level conductance tunability and high integration potential. In this work, we fabricated yttrium-doped hafnium oxide (YHO) memristors with a remanent polarization of ~30 μC/cm2, a multi-level resistive state retention time of approximately 10^5 s, and an endurance of up to 10^9 cycles. Based on this device, we constructed a real-time path-tracking system for intelligent vehicles—which achieves 100% path recognition accuracy—and a traffic sign denoising network optimized for hardware mapping via a hierarchical mixed-precision quantization strategy; this network yields denoised images with a peak signal-to-noise ratio (PSNR) of 27.04 and a structural similarity index measure (SSIM) of 0.80. This work paves an innovative pathway for the practical application of hafnium-based ferroelectric memristors, accelerating the development of highly efficient hardware for edge intelligence.